知识卡片

普通读书笔记卡 · 9174.b · 生成模型:GPT-6

学习卷积核不保证可命名语义

理解等级:中级

多层局部变换可能形成从边缘到复杂模式的层级表示,但某核响应高不证明它只检测一个人类可命名部件。书中全正竖列核对均匀亮区也会响应,并非严格忽略水平结构;解释需要看实际数据与干预。

这张卡片看不懂?

已有反馈 · 0 条

    还没有人提交反馈。

    在整书笔记中的位置

    原书出处 · 1 处

    • · 查看原书依据

      Convolution kernels emphasize areas of similarity. This can be readily understood when considering a simple kernel like [[0,9,0],[0,9,0],[0,9,0]], which emphasizes vertical lines but ignores horizontal ones. Training a convolutional network therefore automatically finds regularities in the training set as well as in the resulting feature map—often generating hierarchical representations by itself. A common example is a convolutional neural network for face detection in which early

    参考来源

    - 位置:《Introduction to Autonomous Robots Mechanisms, Sensors, Actuators, and Algorithms (Nikolaus Correll, Bradley Hayes etc.)》第10章《人工神经网络》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Introduction to Autonomous Robots Mechanisms, Sensors, Actuators, and Algorithms (Nikolaus Correll, Bradley Hayes etc.)/_epub-src/OEBPS/xhtml/chapter_10.xhtml) - 结论依据:摘录中关于学习卷积核不保证可命名语义的定义、比较或条件支撑卡片;原书表述不严谨的范围已在覆盖记录逐项限定,不把案例结果外推为普遍结论。 - 原始内容:Convolution kernels emphasize areas of similarity. This can be readily understood when considering a simple kernel like [[0,9,0],[0,9,0],[0,9,0]], which emphasizes vertical lines but ignores horizontal ones. Training a convolutional network therefore automatically finds regularities in the training set as well as in the resulting feature map—often generating hierarchical representations by itself. A common example is a convolutional neural network for face detection in which early